Dear All,
I'm wondering if there is any way to correctly size the machine required to operates a GRID at both NGI level and site level.
Currently the situation we have (top level):
- WMS (2 machines) are both running on physical machines each having 1 Core Duo 2 CPU and 4GB RAM;
- BDII (2 machines) : one running on a Virtual Machine (4GB RAM and 4 Virtual CPU) and the other on a Core Duo 2 machine with 4 GB RAM;
- VOMS (2 machines): both running as Virtual Machines.
On site level:
- SE and NFS: virtual machines;
- CE and WN running on dedicated hardware :-D.
From my point of view, it seems that virtual VOMS seems OK but BDII not really. Fortunately, doubling machines and load balancing could help. The big advantage of virtual machine is disaster recovery and price per machine.
Now, how could we correctly size machines? I have the impress that BDII will run smoothly on physical servers but how big? Same questions for WMS and VOMS.
What are your experiences, problems encoutered...
The virtual machines are running on a ESX platform for BDII and VOMSes (with enough ressources). The NFS and SE run on a ESXi in order to get access to the private LAN of the WN and being closer.
Thanks for your help.
Kr,
David
David Durvaux
BELNET
Begrid and services
Wetenschapsstraat 4 Rue de la Science
B-1000 Brussel
+32 2 790 33 33
[log in to unmask]
[log in to unmask]
|